Outdoor Blue Spaces, Human Health and Well-Being: a Systematic Review Of

1 Outdoor blue spaces, human health and well-being: a systematic review of

2 quantitative studies

4 Mireia Gascona,b,c, Wilma Zijlemaa,b,c, Cristina Verta,b,c, Mathew P. Whited, Mark

5 Nieuwenhuijsena,b,c

7 Affiliations

8 aISGlobal, Barcelona Ctr. Int. Health Res. (CRESIB), Hospital Clínic - Universitat de

9 Barcelona, Barcelona, Spain

10 bUniversitat Pompeu Fabra (UPF), Barcelona, Spain

11 cCIBER Epidemiología y Salud Pública (CIBERESP), Barcelona, Spain

12 dEuropean Centre for Environment & Human Health, University of Exeter Medical

13 School, Truro, UK.

15 Name and contact of the corresponding author

16 Mireia Gascon

17 Parc de Recerca Biomèdica de Barcelona (PRBB) – ISGlobal (CREAL)

18 Doctor Aiguader, 88 | 08003 Barcelona, Catalonia, Spain

19 Phone: 0034 932147363 Fax: 0034 932045904

20 email: [email protected]

33 Abstract

34 Background: a growing number of quantitative studies have investigated the potential

35 benefits of outdoor blue spaces (lakes, rivers, sea, etc) and human health, but there is

36 not yet a systematic review synthesizing this evidence.

37 Objectives: to systematically review the current quantitative evidence on human health

38 and well-being benefits of outdoor blue spaces.

39 Methods: following PRISMA guidelines for reporting systematic reviews and meta-

40 analysis, observational and experimental quantitative studies focusing on both

41 residential and non-residential outdoor blue space exposure were searched using

42 specific keywords.

43 Results: in total 35 studies were included in the current systematic review, most of them

44 being classified as of “good quality” (N=22). The balance of evidence suggested a

45 positive association between greater exposure to outdoor blue spaces and both benefits

46 to mental health and well-being (N=12 studies) and levels of physical activity (N=13

47 studies). The evidence of an association between outdoor blue space exposure and

48 general health (N= 6 studies), obesity (N=8 studies) and cardiovascular (N=4 studies)

49 and related outcomes was less consistent.

50 Conclusions: although encouraging, there remains relatively few studies and a large

51 degree of heterogeneity in terms of study design, exposure metrics and outcome

52 measures, making synthesis difficult. Further research is needed using longitudinal

53 research and natural experiments, preferably across a broader range of countries, to

54 better understand the causal associations between blue spaces, health and wellbeing.

56 Keywords: outdoor; blue spaces; health; well-being

58 Introduction

59 A growing body of literature suggests that natural outdoor environments might help

60 reduce stress, promote physical activity and social relationships and potentially improve

61 human health and well-being (Dadvand et al., 2016; Hartig et al., 2014; Nieuwenhuijsen

62 et al., 2017; Pasanen et al., 2014; Völker and Kistemann, 2011). Furthermore, the

63 presence of natural environments in urban settings, particularly green spaces (trees,

64 grass, forests, parks, etc), has been associated with a reduction of the levels of air

65 pollution and noise, as well as extreme temperatures in cities, which can lead to a

66 reduction of the impacts of these environmental factors on our health (Burkart et al.,

67 2016; Hartig et al., 2014; Shanahan et al., 2015; Wolf and Robbins, 2015). In a world

68 with rapid urbanization and pressure on space (United Nations Department of Economic

69 and Social Affairs, 2014), evidence of such health benefits of natural environments are

70 of high relevance for healthcare professionals, urban planners and policymakers, who

71 can help translate available evidence into salutogenic interventions and policies (i.e.

72 ones that support and promote health and well-being).

74 Most of the studies in this field to date have focused on the health benefits of green

75 spaces, including several systematic and non-systematic reviews (e.g. Dzhambov et al.,

76 2014; Dzhambov and Dimitrova, 2014; Gascon et al., 2016, 2015; Kabisch et al., 2015;

77 Lee and Maheswaran, 2011; Nieuwenhuijsen et al., 2017; WHO, 2016). However,

78 fewer studies have focused on the health impacts of outdoor blue spaces, defined in the

79 European Commission funded project BlueHealth (https://bluehealth2020.eu/) as

80 “outdoor environments - either natural or manmade - that prominently feature water and

81 are accessible to humans either proximally (being in, on, or near water) or

82 distally/virtually (being able to see, hear or otherwise sense water)” (Grellier et al.,

83 2017). Despite a rich and growing body of multi-faceted, qualitative and narrative work

84 that is shedding important light on people’s interactions with blue space environments,

85 including their perceptions of potential health benefits (Foley and Kistemann, 2015;

86 Völker and Kistemann, 2011), we know less about the quantitative, primarily

87 epidemiological data that seeks to address questions such as “Is living near or having

88 better access to blue space environments associated with positive health and well-being

89 outcomes?”. Although one systematic review, focusing on mortality, did seek to explore

90 the relationship with blue space, in addition to more commonly researched green space,

91 no studies were found at that time to have considered this association (Gascon et al.,

92 2016). However, we know of far more studies that have looked at blue space and

93 morbidity and well-being outcomes although as yet no systematic review of this

94 literature has been conducted. The aim of the current paper was to fill this gap.

95 Moreover, given that it has been hypothesized that any of the potential benefits to health

96 and well-being from exposure to outdoor blue space may follow pathways similar to

97 those identified for green space including stress reduction, increased physical activity,

98 promotion of positive social contacts, increased place attachment and the reduction of

99 extreme temperatures may also play a role for outdoor blue spaces (Grellier et al.,

100 2017), the current review will explore this potential pathways as well.

101

102 Materials and methods

103 Search strategy and selection criteria

104 We followed the PRISMA statement guidelines for reporting systematic reviews and

105 meta-analyses (Moher et al., 2010). The bibliographic search was conducted using

106 MEDLINE (National Library of Medicine) and SCOPUS search engines and keywords

107 related to outdoor blue spaces (blue space, river , lake, sea, beach, fountains, riparian,

108 oceans, coastal, marine) combined with keywords related to health and well-being

109 (mood disorder, depressive disorder, depression, anxiety disorder, anxiety, obsessive-

110 compulsive disorder, stress, mental health, mental hygiene, mental disorders, emotional

111 well-being, psychological well-being, social well-being, well-being, cognitive function,

112 health, cardiovascular diseases, heart diseases, blood pressure, hypertension, obesity,

113 overweigh, weight, body mass index, BMI) as well as keywords related to the potential

114 mechanisms accounting for this association (physical activity, social capital, social

115 contacts, social support). The search was limited to the English language and studies in

116 humans and the last search was conducted on July 1st 2016. All studies published prior

117 to that date and that followed the inclusion criteria were included in the systematic

118 review. In order to have a potentially global overview, the country where the study had

119 been conducted was not an inclusion criterion. Two authors (MG and WZ)

120 independently identified and conducted the first screening of the articles based on the

121 information available in the title and abstract by using the Rayyan tool (Elmagarmid et

122 al., 2014), an online tool that provides procedural support in the selection of articles for

123 systematic reviews. After this first screening stage, based on titles and abstracts, both

124 authors independently read each article to decide whether or not these were eligible for

125 inclusion. After this process, authors compared the final list of papers included by each

126 of them and where disagreements or doubts were encountered the final decision was

127 resolved by discussion between the two independent researchers. We also checked the

128 references of the relevant articles and other sources to find additional articles following

129 the inclusion criteria.

130

131 Study eligibility criteria and quality of the studies

132 The selection criteria were as follows: a) the article needed to be an original research

133 article, b) all type of study designs were included (ecological, prospective, cross-

134 sectional, case-control, and intervention studies), c) the article used quantitative data to

135 report analysis of health and well-being, social or physical activity outcomes in relation

136 to outdoor blue space exposure, d) outdoor blue space exposure was included as a

137 separate variable within the analysis and results were reported specifically for outdoor

138 blue space, even if this was not the primary aim of the study.

139

140 We extracted the following data: author, year of publication, country, study design,

141 study population, sample size, exposure assessment, outcome assessment, confounding

142 factors, and other relevant information including information on potential biases (Tables

143 1-4 and supplemental material Table A). The data extraction work was conducted by the

144 two authors independently, as well as the evaluation of their quality and classification of

145 the evidence. Agreement was reached via consensus. Study quality evaluation was

146 based on an adapted version of the criteria used in previous reviews in the green space

147 field (Gascon et al., 2016, 2015). Further information on how the quality scores were

148 given is provided in Table B of the supplemental material. The following quality scores

149 (%) were derived as follows (supplemental material Table C): excellent quality (score

150 ≥81%), good quality (between 61 and 80%), fair quality (between 41 and 60%), poor

151 quality (between 21 and 40%) and very poor quality (≤20%). These quality scores

152 facilitated establishing the degree of evidence for causal relationships (while

153 recognizing that many of these studies are not designed to test causality directly)

154 between outdoor blue spaces and the outcomes evaluated. The definitions of the degree

155 of evidence for causal relationships used in the current systematic review are an

156 adaptation of the International Agency for Research on Cancer (IARC) definition to

157 define evidence of a causal relationship for cancer (IARC, 2006): sufficient - a positive

158 relationship has been observed between the exposure and outcome in most of the

159 studies, including good quality studies, in which chance, bias and confounding could be

160 ruled out with reasonable confidence; limited - several good quality, independent,

161 studies report a statistically significant association, but chance, bias or confounding

162 cannot be rule out and therefore evidence is not yet conclusive enough; inadequate - if

163 statistically significant associations are reported in one or more studies, but insufficient

164 quality, insufficient number of studies, lack of consistency, and/or lack of statistical

165 power preclude a conclusion regarding the presence or absence of a causal relationship;

166 evidence for lack of association - several good quality studies are consistent in showing

167 no causal relationship. We separately evaluated the evidence according to the different

168 outcomes assessed in the studies included.

169

170 Results

171 We identified 2438 articles in MEDLINE and 2380 articles in SCOPUS and 33 through

172 other sources. After screening the title and the abstracts and checking for duplicates, 91

173 articles were chosen for full-text evaluation, of which 35 articles were finally included

174 in the systematic review as these fulfilled the inclusion criteria (Figure 1).

175

176 Out of the 35 studies included, 29 had a cross-sectional design, of which three were

177 ecological. Five studies were longitudinal and another had a pre-/post-observational

178 study design. The quality of 35 the studies was mainly good (N=22), but one study

179 received a score of excellent quality, six studies were rated as fair and the remaining six

180 studies received a score of poor quality (no study was rated as very poor; supplemental

181 material Table C).

182

183 All studies, except four, were conducted in high-income countries: twenty-one in

184 Europe, including the United Kingdom (n=11), Croatia (N=3), Spain (N=2), France

185 (N=2), The Netherlands (N=1), Finland (N=1), Ireland (N=1). The other studies were

186 conducted in Australia (N=6), China (N=3), New Zealand (N=2), Canada (N=1), the

187 USA (N=1) and Yemen (N=1). Additionally, most studies included in this review were

188 published in the last five years (N=24).

189

190 Outdoor blue space exposure assessment

191 Regarding outdoor blue space exposure assessment we evaluated different aspects:

192 outdoor blue space type (whether the outdoor blue space was inland – rivers, lakes,

193 ponds, streams, rivulets, wetlands, fresh waters - or non-inland – including coast, beach,

194 and salt waters), the environment of assessment (residential, school, leisure, etc.) and

195 the method used to evaluate exposure to outdoor blue spaces (Figure 2). For all three

196 aspects there was a high heterogeneity among studies.

197

198 Inland, non-inland outdoor blue space

199 Most studies only focused on salt water and/or the coast (N=20), while others combined

200 fresh and salt water surfaces or inland and non-inland outdoor blue spaces in the

201 analysis (N=11). Only four were limited to only freshwater (Table 1, 2, 3 and 4).

202

203 Outdoor blue space environment of assessment

204 Most studies (N=28) focused on outdoor blue space exposure around individuals’

205 residential area. Exceptions included three studies evaluating the use of outdoor blue

206 spaces in general (Amoly et al., 2014; Elliott et al., 2015; White et al., 2013c), a study

207 evaluating running routes (Rogerson et al., 2016), a study that focused on exposure to

208 water bodies (e.g. oceans, lakes, rivers, streams) 5 km around schools (in addition to

209 home residence (Huynh et al., 2013), a study that used school location as a proxy for

210 home location (Gilmer et al., 2003), and a time-series study evaluating the real-time use

211 of different spaces, including outdoor blue spaces (MacKerron and Mourato, 2013).

212

213 Method of evaluation

214 Six studies used land-cover or land-use maps and GIS techniques to define the

215 percentage of outdoor blue spaces surface within a defined area (e.g. buffer or census

216 area unit) (Alcock et al., 2015; de Vries et al., 2003; Huynh et al., 2013; Karusisi et al.,

217 2012; Wheeler et al., 2015; White et al., 2013a). Two further studies used the land-

218 cover and land-use maps to define the presence/absence of blue space water

219 environments within a particular buffer (Perchoux et al., 2015; Triguero-Mas et al.,

220 2015). Another common approach, also using land-cover and land-use maps and GIS-

221 techniques, was measuring the distance between the residence, neighbourhood or school

222 of the participants and the edge of outdoor blue space of interest (e.g. coastline); some

223 studies evaluated linear distance (Brereton et al., 2008; Halonen et al., 2014; Wheeler et

224 al., 2012; White et al., 2014, 2013b; Wood et al., 2016; Ying et al., 2015) and others

225 network distance (Edwards et al., 2014; Wilson et al., 2011) or travelling time (Witten

226 et al., 2008). GIS-techniques and land-cover maps were also used to define whether the

227 study subjects lived in a coastal neighborhood (Ball et al., 2007). Three studies defined

228 coastal vs. inland residence based on the postal code (Bauman et al., 1999; Humpel et

229 al., 2004a, 2004b), whereas seven other studies defined exposure to coast based on

230 whether the region of the country or the state had coast or not (Bergovec et al., 2008;

231 Gilmer et al., 2003; Kern et al., 2009; Modesti et al., 2013; Qin et al., 2013; Turek et al.,

232 2001; Zhang et al., 2014). The three studies that used questionnaires as the tool to

233 define outdoor blue space exposure assessed the time spent in blue spaces (Amoly et al.,

234 2014; Elliott et al., 2015; White et al., 2013c), whereas MacKerron and Mourato

235 evaluated real-time exposure to outdoor blue spaces based on a smartphone application

236 with geolocation linked to a land-cover map (MacKerron and Mourato, 2013). Nutsford

237 et al. created an index, named Vertical Visibility Index (VVI), which takes into account

238 different aspects of the built environment to define visibility of outdoor blue spaces (or

239 any other type of natural space) from the residence of the participants, in this case

240 within a buffer of 15 km (Nutsford et al., 2016). Rogerson et al. evaluated exposure to

241 outdoor blue spaces by recruiting runners using the routes that included the beach or the

242 river, and who were therefore considered to be exposed (Rogerson et al., 2016). Finally,

243 Wheeler et al. also factored in freshwater quality, as well as quantity, using an index

244 created by the Environmental Agency for England and Wales (Wheeler et al., 2015).

245

246 Outcomes

247 General health

248 General health was evaluated in six studies (Table 1) (de Vries et al., 2003; Triguero-

249 Mas et al., 2015; Wheeler et al., 2015, 2012; White et al., 2013b; Ying et al., 2015). In

250 four of these studies information on perceived general health and well-being was self-

251 reported by the participants and was based on one single question (which is supposed to

252 have a strong association with more complex dimensions of physical and psychological

253 health) (de Vries et al., 2003; Wheeler et al., 2015, 2012; White et al., 2013b). The

254 study by de Vries et al. additionally asked for the number of symptoms experienced in

255 the last 14 days (a maximum of 43 health symptoms, including e.g. headaches, fatigue,

256 cough) (de Vries et al., 2003). A study conducted in Spain used the Short Form health

257 survey (SF-36) to define perceived general health (Triguero-Mas et al., 2015), whereas

258 a study of the Chinese population asked the question “How do you feel about your

259 health status?” (Ying et al., 2015) (Table 1).

260

261 Results for self-reported general health were heterogeneous; two ecological cross-

262 sectional studies reported better perceived general health in relation to outdoor blue

263 spaces (Wheeler et al., 2015, 2012). The first study observed this association among

264 those living in urban areas less than 1 km from the coast compared to those living more

265 than 50 km away, but not among those living in smaller towns or rural areas (Wheeler et

266 al., 2012). The second study observed beneficial effects among those subjects with a

267 higher percentage of salt and coastal water in their census area. However, no

268 associations were observed with freshwater (Wheeler et al., 2015). A longitudinal study

269 also reported better perceived general health among the same individuals when they

270 lived less than 5 km from the coast compared to years when they lived further inland

271 (White et al., 2013b). However, three cross-sectional studies reported no associations

272 between perceived general health and the percentage of fresh or salt water present

273 within a 3 km buffer (de Vries et al., 2003) or access to inland and non-inland outdoor

274 blue spaces in buffers between 100 m and 1 km (Triguero-Mas et al., 2015; Ying et al.,

275 2015). Although the pioneering de Vries (2003) study did not observe associations with

276 perceived general health, it did observe a reduction in the number of symptoms reported

277 by the participants associated with the percentage of fresh and salt water (de Vries et al.,

278 2003). Finally, Wheeler et al. also evaluated whether ecological quality of freshwater

279 was associated with better perceived general health; however, they observed an inverse

280 association (Wheeler et al., 2015) (Table 1).

281

282 Based on the few studies available, the quality of the studies, the heterogeneity in both

283 study design and results obtained (Table 1), we classified the evidence of an association

284 between outdoor blue space exposure and general health as ‘inadequate’ to make a firm

285 conclusion for either inland or coastal waters at this time.

286

287 Mental health and well-being

288 Mental health and well-being was evaluated in twelve studies (Alcock et al., 2015;

289 Amoly et al., 2014; Brereton et al., 2008; de Vries et al., 2003; Huynh et al., 2013;

290 MacKerron and Mourato, 2013; Nutsford et al., 2016; Rogerson et al., 2016; Triguero-

291 Mas et al., 2015; White et al., 2013b, 2013c, 2013a) (Table 2). Tools to evaluate mental

292 health varied substantially; five studies used the General Health Questionnaire (GHQ)

293 (Alcock et al., 2015; de Vries et al., 2003; Triguero-Mas et al., 2015; White et al.,

294 2013a, 2013b), with remaining studies using a variety of validated questionnaires

295 measuring behavioral and emotional problems, well-being, self-esteem, mood,

296 perceived stress, and psychological distress (Table 2). Six studies used either widely

297 utilized single item measures of well-being e.g. life satisfaction (Brereton et al., 2008;

298 White et al., 2013a, 2013b), happiness (MacKerron and Mourato, 2013) with unknown

299 validity; constructed original scales based on the content of existing surveys e.g.

300 recalled mental restoration (White et al., 2013c) or perceived depression or anxiety, and

301 visits to mental health specialists and medication intake (Triguero-Mas et al., 2015),

302 within untested psychometric properties.

303

304 Three of the five studies using the GHQ to evaluate mental health did not observe

305 associations with the percentage of residential fresh or salt water within the area of

306 study (de Vries et al., 2003; White et al., 2013a) or access to inland and non-inland

307 outdoor blue spaces between 100 m and 1 km (Triguero-Mas et al., 2015). Triguero-

308 Mas et al. also failed to find a significant association with perceived depression or

309 anxiety, visits to mental health specialists or medication intake (tranquilizers,

310 antidepressants, sleeping medication) (Triguero-Mas et al., 2015). However, two other

311 studies, both using data from an 18 year longitudinal survey in the UK, reported

312 positive associations; the first study observed that people’s mental health was better in

313 years when they lived less than 5 km from the coast, compared to years when they were

314 living further inland (White et al., 2013b). The second study, including residents from

315 rural areas, found that while greater exposure to purely coastal areas was again

316 associated with better GHQ scores (over time), quantities of saltwater (estuaries) was

317 associated with worse GHQ scores (over time) (Alcock et al., 2015).These differences

318 were found for people moving between different areas over time, while cross-sectional

319 analyses did not reveal any significant patterns (similarly to most cross-section

320 findings), suggesting that moving location (to ones with more or less blue space) may

321 affect outcomes, but that people may adapt in the long-term so that these differences

322 wash-out over time. White et al. additionally evaluated life satisfaction as a proxy for

323 well-being, but the authors did not observe associations with residential coastal or

324 freshwater distance (White et al., 2013a, 2013b). A third study including life

325 satisfaction reported that those individuals living <2km from the coast were more

326 satisfied with their life than those living >5km. No associations were observed with

327 distances between 2 and 5 km and with residential distance from the beach (Brereton et

328 al., 2008). Nutsford et al. reported a reduction of psychological distress with increasing

329 outdoor blue space (oceanic and freshwater) visibility from the home in a buffer of 15

330 km (Nutsford et al., 2016). Finally, Huyhn et al. used the percentage of water bodies

331 within a 5 km buffer around schools to define exposure to outdoor blue spaces in

332 children aged 11 to 16 years (Huynh et al., 2013). The study observed that increasing

333 exposure to water bodies increased emotional well-being, even after excluding children

334 living outside the 5 km buffer around the schools (Huynh et al., 2013) (Table 2).

335

336 Three studies evaluated the effects of time spent at the beach (Amoly et al., 2014) or

337 any blue space (MacKerron and Mourato, 2013; White et al., 2013c). The first study

338 observed that children (7-10 years) whose parents reported spending more time at the

339 beach during the year had less emotional problems and more pro-social behavior, but no

340 associations were found with Attention Deficit Hyperactivity Disorder (ADHD)

341 symptoms (Amoly et al., 2014). The second study used GPS and a smartphone

342 application for geolocalization, as well as a real-time questionnaire about happiness,

343 which the same individuals completed in many different locations. The results showed

344 that individuals were happier when they reported being in marine and coastal areas, as

345 well as freshwater, wetlands and flood plains than when they were in any other type of

346 urban or rural setting (MacKerron and Mourato, 2013) (Table 2). The third study

347 reported that adolescents and adults spending more time at the beach or the coast had

348 better recalled restoration than urban green spaces (i.e. felt more relaxed/refreshed).

349 However, this association was not statistically significant among those spending more

350 time at inland waters such as rivers, lakes or canals (White et al., 2013c).

351

352 Finally, in a pre-/post-observational study, Rogerson et al. did not observe significant

353 differences in self-esteem, perceived stress or mood between runners running along

354 beach or riverside routes compared to those running along other routes called the

355 “grass” and the “heritage” routes before and after the running (Rogerson et al., 2016)

356 (Table 2).

357

358 Although some good quality studies reported an association between residential and the

359 use of outdoor blue spaces and mental health and well-being, the number of studies is

360 still limited and there is heterogeneity in study design and results among and within

361 them (Table 2). We therefore feel that, based on the classification criteria defined, the

362 most appropriate classification is ‘limited’ for both inland and coastal waters.

363

364 Physical activity

365 Thirteen cross-sectional studies focused on the promotion of physical activity (Ball et

366 al., 2007; Bauman et al., 1999; Edwards et al., 2014; Elliott et al., 2015; Gilmer et al.,

367 2003; Humpel et al., 2004a, 2004b; Karusisi et al., 2012; Perchoux et al., 2015; White et

368 al., 2014; Wilson et al., 2011; Witten et al., 2008; Ying et al., 2015) (Table 3). Six of

369 these studies used non-validated questionnaires, or at least the validation was not

370 reported, whereas the other seven studies used validated tools. Two of these studies

371 used the International Physical Activity Questionnaire (IPAQ) (Ball et al., 2007;

372 Humpel et al., 2004b), another the Youth Health Survey (YHS) (Gilmer et al., 2003),

373 Edwards et al. used the Adolescent Physical Activity Recall Questionnaire (APARQ)

374 (Edwards et al., 2014) and Ying et al. used a pedometer (Ying et al., 2015). Wilson et

375 al. reported previous validation of the questionnaire, but only in a sample of middle-

376 aged women (Wilson et al., 2011), and Bauman et al. also referred to validity and

377 reliability studies for the questionnaire used (Bauman et al., 1999) (Table 3).

378 Additionally, six of the studies defined exposure based on a relatively coarse assessment

379 of proximity to coastal areas (Ball et al., 2007; Bauman et al., 1999; Gilmer et al., 2003;

380 Humpel et al., 2004a, 2004b; Wood et al., 2016), and not on the actual distance between

381 the coastal areas and the address of residence of the participants (Table 3).

382

383 The studies using this relatively coarse exposure assessment obtained diverse results;

384 Ball et al. reported increasing leisure-time and transport walking among adult women

385 living in coastal suburbs (Ball et al., 2007). Bauman et al. also reported a higher

386 likelihood of having a less sedentary lifestyle and achieving an adequate weekly energy

387 expenditure or high levels of physical activity among those adults who lived near the

388 coast (Bauman et al., 1999). However, two other studies including adults did not report

389 associations between living in a coastal area and levels of physical activity (Humpel et

390 al., 2004a, 2004b); only two statistically significant associations were observed in these

391 two studies among the several results reported for men and women separately (Table 3).

392 Finally, a study including youth aged 11-14 years with a parental history of premature

393 coronary heart disease reported statistically significantly lower levels of mean physical

394 activity among youth living in coastal areas compared to non-coastal areas (Gilmer et

395 al., 2003).

396

397 In a French cross-sectional study involving adults from 30 to 79 years of age the

398 percentage of inland water in a buffer of 1 km from the residence was associated with

399 reporting jogging behavior in the previous week (Karusisi et al., 2012). Some years later

400 a similar methodology was applied, but in this case the exposure was defined as the

401 presence of lakes or waterways in different spaces (residential, recreational, work,

402 groceries, social and services) and buffers (Table 3) and the outcome was the reporting

403 of recreational walking in the previous week (Perchoux et al., 2015). The results

404 showed that subjects with access to residential and/or recreational blue spaces were less

405 likely to not report recreation walking in the previous week. However, when outdoor

406 blue spaces from all types of space were included in the analysis no association was

407 observed (Perchoux et al., 2015) (Table 3). In another study distance from the beach,

408 measured as the driving travel time to access the nearest beach, was not associated with

409 meeting the recommended levels of physical activity (at least 2.5 h on five or more days

410 over the preceding week) in adolescents and adults above 15 years of age (Witten et al.,

411 2008). However, those living between 9.2 and 31.8 min from the beach were more

412 likely to report a sedentary behavior profile (< 30 min of physical activity in the past

413 week) compared to those living less than 9.2 min. No associations were observed

414 among those living further than 31.8 min walking distance to the coast (Witten et al.,

415 2008). A study including adults of 40 to 65 years of age in Australia reported that living

416 in a shorter network distance to the river or the coast was associated with spending more

417 time walking for recreation (≥ 300 min/week) (Wilson et al., 2011). However, a similar

418 study in China reported no associations between living within 500m from the river and

419 physical activity measured with a pedometer (Ying et al., 2015). Again in Australia,

420 Edwards et al. observed that rural adolescents reporting the use of the beach for physical

421 activity were more likely to meet physical activity guidelines both in summer and

422 winter, and that those living more than 800m from the beach were less likely to report

423 using the beach for physical activity than those living less than 800m (Edwards et al.,

424 2014). Related to visits to blue spaces and levels of physical activity, a study showed

425 that living <1km from the coast statistically significantly increased the likelihood of

426 doing 30 minutes or more of physical activity at least 5 days a week compared to those

427 living more than 20 km from the coast. The study also showed that this association was

428 mediated by the number of coastal visits and that coastal visits were associated to

429 residential distance from the coast (White et al., 2014). Elliott et al. also reported

430 increased levels of physical activity (MET minutes based on activity and duration of the

431 activity) among those who visited coastal areas (seaside resort and other coastal visits)

432 compared to inland urban and rural green spaces (Elliott et al., 2015).

433

434 Despite the heterogeneity in relation to the study design, mainly in exposure and

435 outcome assessment, most of the studies (N=8/13) were of good quality. Five of these

436 good quality studies reported increasing levels of physical activity with increasing

437 exposure to outdoor blue spaces (Table 3). Nevertheless, despite this relatively

438 consistent association, we classified the evidence for an association as ‘limited’ given

439 the methodological heterogeneities between studies. This classification of the evidence

440 was mainly for non-inland waters, as most of the studies evaluated only beach and

441 coastal areas (N=9/13).

442

443 Obesity and cardiovascular related outcomes

444 Eight studies evaluated obesity related outcomes in relation to blue spaces (Table 4).

445 The only longitudinal study (8 years of follow-up) evaluating the relationship between

446 residential distance to blue spaces (lakes, rivers or sea) and overweight and obesity did

447 not find a significant pattern of associations, either among people who moved home

448 residence over the time or those who did not (Halonen et al., 2014). Two cross-sectional

449 studies comparing adults of coastal and non-coastal areas also failed to find any

450 associations with abdominal obesity (Modesti et al., 2013) or statistically significant

451 differences in BMI between regions (Turek et al., 2001). However, two Chinese studies

452 with a similar study design to those of Modesti et al. and Turek et al. obtained

453 heterogeneous results; Qin et al. reported increased risk of overweight, obesity and

454 abdominal obesity among hypertensive adults aged 45-75 years living inland compared

455 to those living in the coastal county, whereas Zhang et al. reported higher prevalence of

456 abdominal obesity among children (7-18 years) living in coastal districts. Ying et al.

457 found no association between living within 500m distance from a river and BMI

458 (continuous variable) in adults (40-80 years), but did find increased overweight/obesity

459 (as compared to normal weight participants) in this range. In the cross-sectional study of

460 Witten et al., where the estimated driving time to the nearest beach was used as a proxy

461 for the distance to the beach, the BMI (self-reported) of adolescents and adults was

462 higher among those living further (>31.8 min) from the beach compared to those living

463 less than 31.8 min (Witten et al., 2008). A second study, of ecological design, observed

464 that children living less than 1 km from the coast in England had a lower BMI than

465 those living further (more than 20 km) (Wood et al., 2016) (Table 4).

466

467 Four cross-sectional studies also evaluated other outcomes related to cardiovascular

468 disease and diabetes (Table 4). All of these studies based exposure on the region of

469 residence of study subjects (coastal and non-coastal areas). Regarding hypertension, one

470 study reported reduced hypertension among adult coronary heart disease patients in

471 hospitals located in coastal areas (Bergovec et al., 2008). However, another study

472 reported increased risk of hypertension among residents of coastal areas (Modesti et al.,

473 2013) and the other study only reported statistically significant differences in

474 hypertension prevalence in four different regions, but no gradient in distance to the

475 coast (Turek et al., 2001). Finally, Kern et al. defined cardiovascular risk burden

476 (CVRB) based on several cardiovascular factors (Table 4). The authors observed

477 differences between regions, reporting higher prevalence of CVRB among those living

478 in the continental regions compared to those living in the coastal region, particularly

479 among women (Table 4) (Kern et al., 2009).

480

481 No associations or statistically significant differences were reported in relation to

482 cholesterol and triglyceride levels in the two studies that evaluated the relationship with

483 blue spaces (Bergovec et al., 2008; Turek et al., 2001). Only a reduced risk of decreased

484 high-density lipoproteins (HDL) – they transport fat molecules out of artery walls to the

485 liver - was observed in one of the studies (Bergovec et al., 2008). No associations with

486 blue space proximity and diabetes were observed in the two studies that investigated

487 this association (Bergovec et al., 2008; Modesti et al., 2013) (Table 4).

488

489 Given the very limited number of studies examining these outcomes and high

490 heterogeneity in the study design and in the results obtained, we classified the evidence

491 supporting a causal relationship between exposure to blue spaces and cardiovascular

492 disease and diabetes as ‘inadequate’.

493

494 Discussion

495 In the present review we evaluated the current quantitative evidence on the relationship

496 between exposure to outdoor blue space environments such as rivers, lakes and the

497 coast, with a range of health and well-being outcomes. The 35 studies that were selected

498 used a wide range of exposures and outcomes. The body of evidence suggested a

499 positive association between exposure to outdoor blue spaces and mental health and

500 well-being and the promotion of physical activity; however the evidence of any direct

501 causation was ‘limited’. The evidence of an association between outdoor blue space and

502 general health was even weaker, as was the evidence with respect to obesity and

503 cardiovascular related outcomes, in part due to the paucity of studies to date. Further,

504 the review found a large degree of heterogeneity across studies in terms of design,

505 exposures, outcomes assessed and measurement tools, preventing any possibility of

506 conducting meta-analysis.

507

508 Study design

509 Most of the studies included in the review were of cross-sectional design, which are of

510 course unable to rule out reverse causation (e.g. those with better mental health are more

511 likely to move to areas with high residential blue space exposure). Indeed, some of

512 these studies were ecological, a design open to the ecological fallacy, and very few were

513 used longitudinal data. Furthermore, the study evaluating pre- and post- exposure

514 measurement had a number of study design limitations (Rogerson et al., 2016). For

515 instance, study participants were people already running in each of the routes

516 established, and these same people did not run along the other routes (Rogerson et al.,

517 2016). Overall, there is a need for more well-designed longitudinal and intervention

518 studies.

519

520 Outdoor blue spaces exposure assessment and other urban determinants

521 The present review highlights the high heterogeneity among studies considering outdoor

522 blue spaces exposure assessment. In order to facilitate comparability between studies

523 and meta-analyses, efforts to conduct common approaches are recommended in future.

524 To this end we suggest a number of things which could be considered in the future.

525 First, although the main interest of a particular study was usually one type of outdoor

526 blue space (e.g. % of or distance to seawater), other types of outdoor blue spaces could

527 (perhaps even should) be included in the analysis as well (e.g. % inland water), to

528 simultaneously explore multiple blue spaces, and reduce potential confounding across

529 blue space types and aid comparability across more studies. Also, as green spaces are a

530 significant part of natural environments and are often highly correlated with outdoor

531 blue spaces and the outcomes of interest, green spaces should also be included in the

532 analysis (see supplemental material Table A for studies in the current review that did do

533 this), to clarify how the two exposures potentially interact or play the largest role. Other

534 characteristics of blue spaces also need greater consideration in future work, e.g. ease of

535 accessibility to a particular outdoor blue space (e.g. presence of busy roads, high slopes,

536 etc) or degree of urbanity. This would provide more accurate information on the actual

537 potential exposure to a particular outdoor blue space. Another important aspect to

538 consider is the inclusion of the actual use (and the type of uses) of outdoor blue spaces

539 in order to more accurately estimate actual exposure. Finally, one aspect that limits the

540 inference of long-term effects of outdoor blue spaces is the fact that some of the studies

541 did not take into account residential history (e.g. most did not report whether or not

542 respondents had been living in their current home for at least 12 months). On a related

543 point, even in the few longitudinal studies it was assumed that blue spaces were

544 constant over time but there may have been important changes in terms of the

545 development of reservoirs, changes in river courses or coastlines which may have

546 altered proximity for access to blue spaces within a given residential area over time.

547 Again, blue space studies should try and capture these longitudinal changes in

548 exposures when possible although we recognize that this is often difficult (Pearce et al.,

549 2016). Additionally, linkage of (historical) residential information to clinical databases

550 or cohort data provide great opportunities for blue health research and environmental

551 epidemiology in general, but researchers and data providers should be aware of

552 potential privacy issues associated with geographic data (Bovenberg et al., 2016). It

553 should also be acknowledged that many studies were using a relatively coarse

554 assessment of coastal proximity (e.g. living in a coastal area or non-coastal area) and

555 there was also considerable heterogeneity in blue space measurement limiting our

556 ability to make direct comparison across studies or conduct meta-analyses. Future

557 research would benefit from more standardized percentage of land cover buffers and/or

558 network distance parameters.

559

560 Outcome assessment

561 Our literature search resulted in studies focusing on various outcomes: general health,

562 mental health, physical activity, obesity and cardiovascular related outcomes. Many

563 studies included in the present review used non-validated questionnaires, especially

564 studies which focused on physical activity. In this sense, future studies should make use

565 of existing standardized and validated questionnaires, although we recognize that much

566 of this literature is relying on secondary datasets which the authors may not be able to

567 influence in terms of item inclusion. Furthermore, in the case of physical activity,

568 nowadays GIS techniques, apps and smartphones allow researchers to obtain more

569 accurate estimates of both physical activity and exposure through processes of

570 geospatial linkages (Bell et al., 2015; Bort-Roig et al., 2014; Donaire-Gonzalez et al.,

571 2013) or real-time information on different health outcomes, as done by MacKerron et

572 al. (MacKerron and Mourato, 2013). Tools evaluating the use of the outdoor blue spaces

573 and the levels of physical activity conducted around or in these spaces are also an

574 interesting option (McKenzie et al., 2006).

575

576 Mediators

577 As stated in the introduction, stress, physical activity, social contacts and place

578 attachment, and the reduction of extreme temperatures, are potential pathways of the

579 association between outdoor blue spaces and health and well-being indicators. In the

580 present review only one article attempted to examine mediation (i.e. visit frequency as a

581 function of coastal proximity, White et al. 2014), and thus far more work is needed in

582 order to understand whether certain health determinants could explain the associations

583 observed between outdoor blue spaces and health. Indeed, a recent systematic review on

584 green spaces and mental health highlighted that some of the included studies observed

585 that the health benefits of green spaces could be mediated by some of these variables

586 (Gascon et al., 2015). Including these analyses would help understand the mechanisms

587 of the association between outdoor blue spaces and health.

588

589 Effect modifiers

590 It is also necessary to identify specific populations that could benefit more from urban

591 planning actions (i.e. neighbourhoods of lower socioeconomic positions, women,

592 children, elderly, etc.), in order to reduce health inequalities (WHO, 2012). In fact,

593 Wheeler et al. conducted such analyses and observed that beneficial effects of outdoor

594 blue spaces were largest in lower socioeconomic areas (Wheeler et al., 2015, 2012) but

595 none of the other studies in this review explored this issue.

596

597 Limitations of this review

598 Although we believe this is the first systematic review in this area and we endeavored to

599 use standard reviewing and grading protocols, we also recognize several limitations.

600 First, the concept of “blue space” has not been widely used compared to, for instance,

601 green space. In the present review we used search terms based on the keywords and title

602 words used in the articles that we already knew about on the topic. However, at the end,

603 and despite the efforts to find all relevant articles, only 18 of the 35 articles included in

604 the review were found through the traditional online search databases; the rest were

605 obtained through other means such as searching citations of included papers, which may

606 have resulted in some key papers being missed. The causes for not detecting half of the

607 papers through the standard research tools could be that: 1) the blue spaces concept, in

608 contrast with green spaces, is not a common concept; or that 2) some studies were not

609 directly designed to evaluate health impacts of outdoor “blue exposure” but rather of

610 several urban characteristics, and did not focus on these relationships in titles or

611 abstracts despite having examined them. In this sense, the BlueHealth project aims to

612 explore this concept and to provide recognition and importance to these spaces, both in

613 research and in the implementation of policies to promote and protect public health. We

614 also recognize that scoring the quality of the papers and classifying the evidence is

615 associated with a degree of subjectivity (e.g. “was there careful consideration of

616 confounders” is open to interpretation). In the present review, in order to reduce

617 subjectivity and bias within the review process itself, two independent reviewers

618 conducted the quality appraisal of each paper independently and disagreements

619 discussed between the two reviewers, but again some bias may still have occurred. Even

620 if this was the case, however, we do not feel that the overall conclusions from our

621 review would substantially change if certain studies were reclassified as from fair to

622 poor or good and vice versa, given the few number of available studies and the

623 heterogeneity among them regarding study design. We also of course recognize the

624 possibility of publication bias, which may be inflating the number of positive

625 associations in the extant literature.

626

627 Other aspects to be considered

628 For this systematic review we decided to exclude qualitative research, as we aimed to

629 focus on those studies quantifying the health impacts of blue spaces. However, we still

630 think it is important to recognize the existing qualitative evidence in this field of

631 research, which already suggests a range of benefits of being exposed to blue spaces

632 (Finlay et al., 2015; Foley, 2015; Foley and Kistemann, 2015; Kearns et al., 2015;

633 Völker and Kistemann, 2015, 2011), and indeed, addresses aspects that quantitative

634 research often cannot or does not evaluate, such as perception, sensory exposure, or

635 experiences that could explain the associations found by those studies conducting

636 quantitative research. Moreover, this type of research is very relevant and should also be

637 considered by policy makers when conducting interventions in urban or natural

638 environments involving blue spaces.

639

640 In the current systematic review none of the studies addressed potential negative

641 impacts of blue spaces or aspects related to blue spaces (e.g. drowning, pollution,

642 flooding, etc). Far more work has been conducted on these issues already, which is why

643 we focus on the benefits here, but there is no doubt that when a more complete picture

644 of the benefits emerges it will be time to conduct the kind of full cost-benefit analyses

645 usually required by policy makers (Grellier et al., 2017).

646

647 Finally, the relationship that people establish with blue spaces can be different across

648 cultures and climates worldwide, and therefore blue spaces may have different

649 meanings, functions and impacts on health and well-being. In the current review we

650 included studies from different countries, but studies from the United Kingdom and

651 Australia represented a large part of the included studies (49%). Therefore, our

652 conclusions must be tentative at this point given the possibility of inherent “cultural

653 bias”. In this sense, international projects like BlueHealth, in which the same

654 approaches and methodologies are applied in many countries or geographical areas at

655 once, may help to better understand any cultural similarities or differences, and what

656 role climatic factors may play in these issues.

657

658 Conclusion

659 As far as we are aware, this is the first systematic review of quantitative studies on

660 outdoor blue spaces and human health and well-being. Results indicate this is emerging

661 topic, as most of the studies have been conducted in the last five years. Also, results

662 suggest potential health benefits of outdoor blue spaces, mainly with respect to mental

663 health and well-being and the promotion of physical activity. However, further studies,

664 with methodological improvements, standardized exposure and outcome measures

665 enabling comparison across cultures and climates, and including urban health

666 determinants other than outdoor blue spaces are needed in order to advance in our

667 knowledge on the topic. Despite the limitations, current scientific evidence supports the

668 promotion and recovery of outdoor blue spaces within urban planning as an interesting

669 strategy for the promotion of health and well-being.

670

671 Acknowledgements

672 ISGlobal is a member of the CERCA Programme, Generalitat de Catalunya. The

673 authors would like to thank Dieneke Schram, Jeroen Van Leuken and Tanja Wolf for

674 their comments and advice.

675

676 Funding

677 This project has received funding from the European Union’s Horizon 2020 research

678 and innovation programme under grant agreement No 666773. The authors have no

679 conflicts of interest.

680 References

681 Alcock, I., White, M.P., Lovell, R., Higgins, S.L., Osborne, N.J., Husk, K., Wheeler, 682 B.W., 2015. What accounts for “England”s green and pleasant land’? A panel data 683 analysis of mental health and land cover types in rural England. Landsc. Urban 684 Plan. 142, 38–46. 685 Amoly, E., Dadvand, P., Forns, J., López-Vicente, M., Basagaña, X., Julvez, J., 686 Alvarez-Pedrerol, M., Nieuwenhuijsen, M.J., Sunyer, J., 2014. Green and Blue 687 Spaces and Behavioral Development in Barcelona Schoolchildren: The BREATHE 688 Project. Environ. Health Perspect. 689 Ball, K., Timperio, A., Salmon, J., Giles-Corti, B., Roberts, R., Crawford, D., 2007. 690 Personal, social and environmental determinants of educational inequalities in 691 walking: a multilevel study. J. Epidemiol. Community Health 61, 108–14. 692 Bauman, A., Smith, B., Stoker, L., Bellew, B., Booth, M., 1999. Geographical 693 influences upon physical activity participation: evidence of a “coastal effect”. Aust. 694 N. Z. J. Public Health 23, 322–4. 695 Bell, S.L., Phoenix, C., Lovell, R., Wheeler, B.W., 2015. Seeking everyday wellbeing: 696 The coast as a therapeutic landscape. Soc. Sci. Med. 142. 697 Bergovec, M., Reiner, Z., Milicić, D., Vrazić, H., 2008. Differences in risk factors for 698 coronary heart disease in patients from continental and Mediterranean regions of 699 Croatia. Wien. Klin. Wochenschr. 120, 684–92. 700 Bort-Roig, J., Gilson, N.D., Puig-Ribera, A., Contreras, R.S., Trost, S.G., 2014. 701 Measuring and influencing physical activity with smartphone technology: a 702 systematic review. Sports Med. 44, 671–86. 703 Bovenberg, J.A., Knoppers, B.M., Hansell, A., de Hoogh, K., 2016. Exposing 704 participants? Population biobanks go geo. Eur. J. Hum. Genet. 24, 155–6. 705 Brereton, F., Clinch, J.P., Ferreira, S., 2008. Happiness, geography and the 706 environment. Ecol. Econ. 65, 386–396. 707 Burkart, K., Meier, F., Schneider, A., Breitner, S., Canário, P., Alcoforado, M.J., 708 Scherer, D., Endlicher, W., 2016. Modification of Heat-Related Mortality in an 709 Elderly Urban Population by Vegetation (Urban Green) and Proximity to Water 710 (Urban Blue): Evidence from Lisbon, Portugal. Environ. Health Perspect. 124, 711 927–934. 712 Dadvand, P., Bartoll, X., Basagaña, X., Dalmau-Bueno, A., Martinez, D., Ambros, A., 713 Cirach, M., Triguero-Mas, M., Gascon, M., Borrell, C., Nieuwenhuijsen, M.J., 714 2016. Green spaces and General Health: Roles of mental health status, social 715 support, and physical activity. Environ. Int. 91, 161–167. 716 de Vries, S., Verheij, R.A., Groenewegen, P.P., Spreeuwenberg, P., 2003. Natural 717 environments -- healthy environments? An exploratory analysis of the relationship 718 between greenspace and health. Environ. Plan. A 35, 1717–1731. 719 Donaire-Gonzalez, D., de Nazelle, A., Seto, E., Mendez, M., Nieuwenhuijsen, M.J., 720 Jerrett, M., 2013. Comparison of physical activity measures using mobile phone- 721 based CalFit and Actigraph. J. Med. Internet Res. 15, e111. doi:10.2196/jmir.2470 722 Dzhambov, A.M., Dimitrova, D.D., 2014. Urban green spaces’ effectiveness as a 723 psychological buffer for the negative health impact of noise pollution: a systematic 724 review. Noise Health 16, 157–65. 725 Dzhambov, A.M., Dimitrova, D.D., Dimitrakova, E.D., 2014. Association between 726 residential greenness and birth weight: Systematic review and meta-analysis. 727 Urban For. Urban Green. 13. 728 Edwards, N.J., Giles-Corti, B., Larson, A., Beesley, B., 2014. The Effect of Proximity

729 on Park and Beach Use and Physical Activity Among Rural Adolescents. J. Phys. 730 Act. Heal. 11, 977–984. 731 Elliott, L.R., White, M.P., Taylor, A.H., Herbert, S., 2015. Energy expenditure on 732 recreational visits to different natural environments. Soc. Sci. Med. 139, 53–60. 733 Elmagarmid, A., Fedorowicz, Z., Hammady, H., Ilyas, I., Khabsa, M., Ouzzani, M., 734 2014. Rayyan: a systematic reviews web app for exploring and filtering searches 735 for eligible studies for Cochrane Reviews, in: Evidence-Informed Public Health: 736 Opportunities and Challenges. Abstracts of the 22nd Cochrane Colloquium. John 737 Wiley & Sons, Hyderabad, India, India. 738 Finlay, J., Franke, T., McKay, H., Sims-Gould, J., 2015. Therapeutic landscapes and 739 wellbeing in later life: Impacts of blue and green spaces for older adults. Health 740 Place 34, 97–106. 741 Foley, R., 2015. Swimming in Ireland: Immersions in therapeutic blue space. Health 742 Place 35, 218–25. 743 Foley, R., Kistemann, T., 2015. Blue space geographies: Enabling health in place. 744 Health Place 35, 157–65. 745 Gascon, M., Triguero-Mas, M., Martínez, D., Dadvand, P., Forns, J., Plasència, A., 746 Nieuwenhuijsen, M., 2015. Mental Health Benefits of Long-Term Exposure to 747 Residential Green and Blue Spaces: A Systematic Review. Int. J. Environ. Res. 748 Public Health. 749 Gascon, M., Triguero-Mas, M., Martínez, D., Dadvand, P., Rojas-Rueda, D., Plasència, 750 A., Nieuwenhuijsen, M.J., 2016. Residential green spaces and mortality: A 751 systematic review. Environ. Int. 86, 60–7. 752 Gilmer, M.J., Harrell, J.S., Miles, M.S., Hepworth, J.T., 2003. Youth characteristics and 753 contextual variables influencing physical activity in young adolescents of parents 754 with premature coronary heart disease. J. Pediatr. Nurs. 18, 159–68. 755 Grellier, J., White, M.P., Albin, M., Bell, S., Elliott, L.R., Gascón, M., Gualdi, S., 756 Mancini, L., Nieuwenhuijsen, M.J., Sarigiannis, D.A., van den Bosch, M., Wolf, 757 T., Wuijts, S., Fleming, L.E., 2017. BlueHealth: a study programme protocol for 758 mapping and quantifying the potential benefits to public health and well-being 759 from Europe’s blue spaces. BMJ Open 7, e016188. 760 Halonen, J.I., Kivimäki, M., Pentti, J., Stenholm, S., Kawachi, I., Subramanian, S. V, 761 Vahtera, J., 2014. Green and blue areas as predictors of overweight and obesity in 762 an 8-year follow-up study. Obesity (Silver Spring). 22, 1910–7. 763 Hartig, T., Mitchell, R., de Vries, S., Frumkin, H., 2014. Nature and health. Annu. Rev. 764 Public Health 35, 207–28. 765 Humpel, N., Owen, N., Iverson, D., Leslie, E., Bauman, A., 2004a. Perceived 766 environment attributes, residential location, and walking for particular purposes. 767 Am. J. Prev. Med. 26, 119–25. 768 Humpel, N., Owen, N., Leslie, E., Marshall, A.L., Bauman, A.E., Sallis, J.F., 2004b. 769 Associations of location and perceived environmental attributes with walking in 770 neighborhoods. Am. J. Health Promot. 18, 239–42. 771 Huynh, Q., Craig, W., Janssen, I., Pickett, W., 2013. Exposure to public natural space as 772 a protective factor for emotional well-being among young people in Canada. BMC 773 Public Health 13, 407. 774 IARC, 2006. (International Agency for Research on Cancer) Preamble to the IARC 775 Monographs (amended January 2006). Scientific review and evaluation. [WWW 776 Document]. URL 777 http://monographs.iarc.fr/ENG/Preamble/currentb6evalrationale0706.php 778 Kabisch, N., Qureshi, S., Haase, D., 2015. Human–environment interactions in urban

779 green spaces — A systematic review of contemporary issues and prospects for 780 future research. Environ. Impact Assess. Rev. 50, 25–34. 781 Karusisi, N., Bean, K., Oppert, J.-M., Pannier, B., Chaix, B., 2012. Multiple dimensions 782 of residential environments, neighborhood experiences, and jogging behavior in 783 the RECORD Study. Prev. Med. (Baltim). 55, 50–5. 784 Kearns, R.A., Collins, D., Conradson, D., 2015. Reprint of: A healthy island blue space: 785 From space of detention to site of sanctuary. Health Place 35, 178–86. 786 Kern, J., Polasek, O., Milanović, S.M., Dzakula, A., Fister, K., Strnad, M., Ivanković, 787 D., Vuletić, S., 2009. Regional pattern of cardiovascular risk burden in Croatia. 788 Coll. Antropol. 33 Suppl 1, 11–7. 789 Lee, A.C.K., Maheswaran, R., 2011. The health benefits of urban green spaces: a review 790 of the evidence. J. Public Health (Oxf). 33, 212–22. 791 MacKerron, G., Mourato, S., 2013. Happiness is greater in natural environments. Glob. 792 Environ. Chang. 23, 992–1000. 793 McKenzie, T.L., Cohen, D.A., Sehgal, A., Williamson, S., Golinelli, D., 2006. System 794 for Observing Play and Recreation in Communities (SOPARC): Reliability and 795 Feasibility Measures. J. Phys. Act. Health 3 Suppl 1, S208–S222. 796 Modesti, P.A., Bamoshmoosh, M., Rapi, S., Massetti, L., Al-Hidabi, D., Al Goshae, H., 797 2013. Epidemiology of hypertension in Yemen: effects of urbanization and 798 geographical area. Hypertens. Res. 36, 711–7. 799 Moher, D., Liberati, A., Tetzlaff, J., Altman, D.G., 2010. Preferred reporting items for 800 systematic reviews and meta-analyses: the PRISMA statement. Int. J. Surg. 8, 336– 801 41. 802 Nieuwenhuijsen, M.J., Khreis, H., Triguero-Mas, M., Gascon, M., Dadvand, P., 2017. 803 Fifty Shades of Green: Pathway to Healthy Urban Living. Epidemiology 28, 63– 804 71. 805 Nutsford, D., Pearson, A.L., Kingham, S., Reitsma, F., 2016. Residential exposure to 806 visible blue space (but not green space) associated with lower psychological 807 distress in a capital city. Health Place 39, 70–78. 808 Pasanen, T.P., Tyrväinen, L., Korpela, K.M., 2014. The relationship between perceived 809 health and physical activity indoors, outdoors in built environments, and outdoors 810 in nature. Appl. Psychol. Health Well. Being. 6, 324–46. 811 Pearce, J., Shortt, N., Rind, E., Mitchell, R., 2016. Life Course, Green Space and 812 Health: Incorporating Place into Life Course Epidemiology. Int. J. Environ. Res. 813 Public Health 13. doi:10.3390/ijerph13030331 814 Perchoux, C., Kestens, Y., Brondeel, R., Chaix, B., 2015. Accounting for the daily 815 locations visited in the study of the built environment correlates of recreational 816 walking (the RECORD Cohort Study). Prev. Med. (Baltim). 81, 142–149. 817 Qin, X., Zhang, Y., Cai, Y., He, M., Sun, L., Fu, J., Li, J., Wang, B., Xing, H., Tang, G., 818 Wang, X., Xu, X., Xu, X., Huo, Y., 2013. Prevalence of obesity, abdominal obesity 819 and associated factors in hypertensive adults aged 45-75 years. Clin. Nutr. 32, 820 361–7. 821 Rogerson, M., Brown, D.K., Sandercock, G., Wooller, J.-J., Barton, J., 2016. A 822 comparison of four typical green exercise environments and prediction of 823 psychological health outcomes. Perspect. Public Health 136, 171–80. 824 Shanahan, D.F., Lin, B.B., Bush, R., Gaston, K.J., Dean, J.H., Barber, E., Fuller, R.A., 825 2015. Toward Improved Public Health Outcomes From Urban Nature. Am. J. 826 Public Health e1–e8. 827 Triguero-Mas, M., Dadvand, P., Cirach, M., Martínez, D., Medina, A., Mompart, A., 828 Basagaña, X., Gražulevičienė, R., Nieuwenhuijsen, M.J., 2015. Natural outdoor

829 environments and mental and physical health: Relationships and mechanisms. 830 Environ. Int. 77, 35–41. doi:10.1016/j.envint.2015.01.012 831 Turek, S., Rudan, I., Smolej-Narancić, N., Szirovicza, L., Cubrilo-Turek, M., Zerjavić- 832 Hrabak, V., Rak-Kaić, A., Vrhovski-Hebrang, D., Prebeg, Z., Ljubicić, M., 833 Janićijević, B., Rudan, P., 2001. A large cross-sectional study of health attitudes, 834 knowledge, behaviour and risks in the post-war Croatian population (the First 835 Croatian Health Project). Coll. Antropol. 25, 77–96. 836 United Nations Department of Economic and Social Affairs, 2014. World Urbanization 837 Prospects - The 2014 Revision - Highlights. United Nations, New York. 838 Völker, S., Kistemann, T., 2015. Developing the urban blue: Comparative health 839 responses to blue and green urban open spaces in Germany. Health Place 35, 196– 840 205. 841 Völker, S., Kistemann, T., 2011. The impact of blue space on human health and well- 842 being - Salutogenetic health effects of inland surface waters: a review. Int. J. Hyg. 843 Environ. Health 214, 449–60. 844 Wheeler, B.W., Lovell, R., Higgins, S.L., White, M.P., Alcock, I., Osborne, N.J., Husk, 845 K., Sabel, C.E., Depledge, M.H., 2015. Beyond greenspace: an ecological study of 846 population general health and indicators of natural environment type and quality. 847 Int. J. Health Geogr. 14, 17. 848 Wheeler, B.W., White, M., Stahl-Timmins, W., Depledge, M.H., 2012. Does living by 849 the coast improve health and wellbeing? Health Place 18, 1198–201. 850 White, M.P., Alcock, I., Wheeler, B.W., Depledge, M.H., 2013a. Would you be happier 851 living in a greener urban area? A fixed-effects analysis of panel data. Psychol. Sci. 852 24, 920–8. 853 White, M.P., Alcock, I., Wheeler, B.W., Depledge, M.H., 2013b. Coastal proximity, 854 health and well-being: results from a longitudinal panel survey. Health Place 23, 855 97–103. 856 White, M.P., Pahl, S., Ashbully, K., Herber, S., Depledge, M.H., 2013c. Feelings of 857 restoration from recent nature visits. J. Environ. Psychol. 35, 40–51. 858 White, M.P., Wheeler, B.W., Herbert, S., Alcock, I., Depledge, M.H., 2014. Coastal 859 proximity and physical activity: Is the coast an under-appreciated public health 860 resource? Prev. Med. (Baltim). 69, 135–40. 861 WHO, 2016. Urban Green space and health: a review of the evidence. Copenhagen. 862 WHO, 2012. Environmental health inequalities in Europe. Assessment report. WHO 863 European Centre for Environment and Health, Bonn. 864 Wilson, L.-A.M., Giles-Corti, B., Burton, N.W., Giskes, K., Haynes, M., Turrell, G., 865 2011. The association between objectively measured neighborhood features and 866 walking in middle-aged adults. Am. J. Health Promot. 25, e12-21. 867 Witten, K., Hiscock, R., Pearce, J., Blakely, T., 2008. Neighbourhood access to open 868 spaces and the physical activity of residents: a national study. Prev. Med. (Baltim). 869 47, 299–303. 870 Wolf, K.L., Robbins, A.S.T., 2015. Metro nature, environmental health, and economic 871 value. Environ. Health Perspect. 123, 390–8. 872 Wood, S.L., Demougin, P.R., Higgins, S., Husk, K., Wheeler, B.W., White, M., 2016. 873 Exploring the relationship between childhood obesity and proximity to the coast: A 874 rural/urban perspective. Health Place 40, 129–36. 875 Ying, Z., Ning, L.D., Xin, L., 2015. Relationship Between Built Environment, Physical 876 Activity, Adiposity, and Health in Adults Aged 46-80 in Shanghai, China. J. Phys. 877 Act. Health 12, 569–78. 878 Zhang, Y., Zhao, J., Chu, Z., Tan, H., 2014. Prevalence and regional disparities in

879 abdominal obesity among children and adolescents in Shandong, China, surveyed 880 in 2010. Ann. Nutr. Metab. 64, 137–43. 881 882

883 Table 1. Main characteristics and results of the studies on general health. Author Study design Age of the N Blue space Tool to estimate blue Outcome measured Tools to measure Main results (year, study exposure space exposure outcome country) population De Vries et Cross-sectional All ages 10197 Fresh and salt Land-use map % BS in Perceived general Questionnaire a β=-0.008 (SD=0.005) (for 3km buffer) al. 2003, The water surface 1km & 3km buffers from health (less than good) Netherlands the residence Number of symptoms Questionnaire β=-0.016 (SD=0.007; significant experienced in the last (43 symptoms)a association) (for 3km buffer) 14 days Triguero- Cross-sectional 34-64 years 8793 Inland and non- BS presence in 100m, Perceived general SF-36 OR (95%CI) = 0.82 (0.65, 1.05) for 300 m Mas et al. inland BS 300m, 500m and 1km health (poor) buffer 2015, Spain buffers from the residence (land-cover) Wheeler et Ecological Not specified 32482 Coast GIS to measure distance Self-reported good Questionnairea β (95%CI) for distance < 1km vs > 50 km = la. 2012, UK Cross-sectional CAUs between population- health in the last 12 m 1.13 (0.99, 1.27) in urban settings, 1.19 weighted centroid and the (0.79, 1.59) in town settings and -0.09 (- coast (> 50 km, 20-50 km, 0.69, 0.51) in rural settings 5-20 km, 1-5 km, < 1km) Wheeler et Ecological Not specified 31672 Salt water, Land-cover map to measure Self-reported good Questionnairea β (95%CI) = 0.074 (0.032, 0.117) for la. 2015, UK Cross-sectional CAUs freshwater, % of BS within each health saltwater, 0.000 (-0.010, 0.010 for coastal population-weighted freshwater and 0.019 (0.010, 0.027) for centroid coastal water Freshwater Indicator of the β (95%CI) = -0.095 (-0.153, -0.037) ecological quality Environment agency for England and Wales map White et al. Longitudinal Adults > 25 Between Coast Distance between General health (good) Questionnairea β (SE; *p-value<0.05) = 0.039 (0.018)* for 2013a, UK years 12360 and population-weighted < 5 km and -0.005 (0.018) for > 50 km 15471b centroid and the coast (< 5 km, 5-50 km (reference category), > 50 km) Ying et al. Cross-sectional Adults 46-80 1100 River GIS to define 500m buffer Health status (worst to Questionnairea Coefficient (p-value)=-0.016 (0.679) 2015, China years distance to the river best)

884 BS: blue space; CAU level: Census Area Unit level; GIS: geographic information system; SF-36: Short form health survey (36 items). 885 aNot validated questionnaire or not clarified 886 bIn total there were between 90084 and 136756 observations included in the analysis 887 cIn total there were between 56574 and 87573 observations included in the analysis 888

889 Table 2. Main characteristics and results of the studies on mental health and well-being. Author (year, Study design Age of the study N Blue space Tool to estimate blue Outcome measured Tools to measure Main results country) population exposure space exposure outcome Alcock et al. Longitudinal (18 Adults (rural 2020a Saltwater, Land-cover map to Mental health GHQ-12 β (SE; *p-value<0.05) = 0.3398 (0.7027) 2015, UK years follow-up) areas) freshwater, determine % at CAU for saltwater, -0.289 (0.0693) for coastal freshwater, 0.298 (0.0818) for coastal between-individuals -2.4219 (1.1305*) for saltwater, 0.4700 (0.3545) for freshwater and 0.8380 (0.3318*) for coastal within-individuals Amoly et al. Cross-sectional Children 7-10 2111 Annual Questionnaire Emotional & SDQ Percent change (95% CI) for an IQR 2014, Spain years beach behavioural problems ADHD/DSM-IV increase in annual beach attendance (32 attendance days) = (time spent) –3.9 (–7.2, –0.4) for SDQ, 1.1 (0.0, 2.2) for pro-social behaviour and –0.1 (–6.7, 6.9) for ADHD Brerenton et al. Cross-sectional Adults >18 years 1467 Proximity Land-use map based on Well-being/life Questionnaireb β (SE; *p-value<0.05) = 0.8351 (4.32*) 2008, Ireland to coast and GIS information to satisfaction for <2km from the coast and 0.2271 (1.51) beach determine proximity to for 2-5km from the coast coast (<2km, 2-5km, -0.1607 (0.73) for <5km from the beach >5km) and to beach and -0.0923 (0.42) for 5-10km from the (<5km, 5-10, >10km) beach De Vries et al. Cross-sectional All ages 10197 Fresh and Land-use map % BS in Minor psychiatric GHQ β=-0.004 (SD=0.005) (for 3km buffer) 2003, The salt water 1km & 3km buffers from morbidity Netherlands surface the residence Huynh et al. Cross-sectional Children 11-16 17249 Water Land-use map based on (Positive) emotional Cantril’s ladder RR (95%CI) =1.02 (0.97, 1.07) for Q2 of 2013, Canada years bodies GIS information. Water well-being BS exposure, 1.06 (1.01, 1.11) for Q3 and (oceans, bodies (%) in 5 km buffer 1.04 (0.99, 1.09) for Q4c lakes, from the schools (N=371) rivers, streams) MacKerron et Longitudinal Adults 21947d Marine and Smartphone application Happiness Smartphone application β (SE, *p<0.05)=6.02 (0.68*) for marina al. 2013, UK coastal with GPS location linked with questionnaireb and coastal margins and 1.80 (0.63*) for margins and to land cover map freshwater, wetlands and flood plains freshwater, wetlands and flood plains

Author (year, Study design Age of the study N Blue space Tool to estimate blue Outcome measured Tools to measure Main results country) population exposure space exposure outcome Nutsford et al. Cross-sectional Adolescents and 442 Oceanic and VVI of BS created based Psychological distress K10 β (95%CI) = -0.28 (-0.41, -0.15) (increase 2016, New adults >15 years freshwater on 4 land-cover maps and of 10% of BS visibility, 0.28% reduction Zealand blue spaces for a buffer of 15 km of psychological stress) using residential meshblocks

Rogerson et al. Pre/post Adult park 331 Beach, Each participant run along Self-esteem RSE (pre/post running There were no differences pre- and post- 2016, UK observational study runners riverside, one of the routes (they at each run between the different locations grass and were recruited at the start location) (p>0.05) heritage of the route) Perceived stress PSS (pre/post running There were no differences pre- and post- routes at each location) run between the different locations (p>0.05) Mood POMS (pre/post running There were no differences pre- and post- at each location) run between the different locations (p>0.05) Triguero-Mas et Cross-sectional 34-64 years 8793 Inland and BS presence in 100m, Perceived mental GHQ-12 OR (95%CI) = 1.13 (0.86, 1.49) for 300 m al. 2015, Spain non-inland 300m, 500m and 1km health (poor) buffer BS buffers from the residence (land-cover) Perceived depression Questionnaireb OR (95%CI) = 1.13 (0.90, 1.41) for 300 m or anxiety (yes/no) buffer Visits to mental health Questionnaireb OR (95%CI) = 1.30 (0.92, 1.84) for 300 m specialists (yes/no) buffer Medication intake Questionnaireb OR (95%CI) = 0.85 (0.61, 1.17) (yes/no) tranquilizers, 0.84 (0.59, 1.19) for antidepressants and 0.95 (0.69, 1.31) for sleeping medication for a 300m buffer White et al. Longitudinal Adults > 25 years Between Coast Distance between Mental health GHQ-12 β (SE; *p-value<0.05) = 0.147 (0.065)* 2013a, UK 12360 and population-weighted for < 5 km and -0.121 (0.064) for > 50 km 15471e centroid and the coast (< 5 km, 5-50 km (reference Well-being (life Questionnaireb β (SE; *p-value<0.05) = 0.044 (0.034) for category), > 50 km) satisfaction) < 5 km and 0.049 (0.034) for > 50 km White et al. Longitudinal Adults Between Freshwater Land-cover map to Mental health GHQ-12 β (SE) = -0.0007 (0.0028) (p-value > 0.05) 2013b, UK 10168 and determine % of 12818f freshwater at CAU Well-being (life Questionnaireb β (SE) = 0.0018 (0.0014) (p-value > 0.05) satisfaction)

Author (year, Study design Age of the study N Blue space Tool to estimate blue Outcome measured Tools to measure Main results country) population exposure space exposure outcome White et al. Cross-sectional Adolescents and 4255 Coast, Questionnaire to ask for Recalled restoration Questionnaireb β (SE; *p-value<0.05) = 0.13 (0.07) for 2013c, UK adults > 16 years beach, river, the time spent in these BS river/lake/canal, 0.21 (0.07)* for beach, lake, canal 0.19 (0.08)* for coast (other than beach)

890 ADHD/DSM-IV: Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition; BS: blue space; CAU level: Census area unit level; GHQ-(12): General Health Questionnaire-(12 891 items); GIS: geographic information system; IQR: Interquartile range; K10: Kessler Psychological Distress Scale (10 items); POMS: Profile of Mood States; PSS: Perceived Stress Scale; RSE: 892 Rosenberg Self-esteem Scale; SDQ: Strengths and Difficulties Questionnaire; VVI: Vertical Visibility Index. 893 aWith 12697 observations included in the analysis 894 bNot validated questionnaire or not clarified if the questionnaire is validated 895 cResults are similar in a sensitive analysis including only those living within the 5km buffer. 896 dWith 1138481 observations included in the analysis 897 eIn total there were between 90084 and 136756 observations included in the analysis 898 fIn total there were between 56574 and 87573 observations included in the analysis 899 900 901 902

903 Table 3. Main characteristics and results of the studies on physical activity. Author Study design Age of the study N Blue space Tool to estimate blue space Outcome measured Tools to measure Main results (year, population exposure exposure outcome country) Ball et al. Cross-sectional Women18-65 1282 Coastal Living (yes/no) in a coastal Walking for leisure- IPAQ (long version) OR (95%CI) = 1.46 (1.02, 1.90) for 2007, years neighbourhood (bayside) suburb defined with time and transport leisure-time walking and 2.74 (2.20, 3.28) Australia GIS and Open Space 2002 for transport walking Bauman et Cross-sectional Adults 16178 Coast Use of the postal code to define Physical activity Questionnaire (to OR (95%CI) = 0.77 (0.69, 0.87) for being al. 1999, coastal residence vs inland measure energy sedentary, 1.27 (1.18, 1.37) for being Australia residence expenditure) adequately active, 1.38 (1.25, 1.52) for having a high level of activity Edwards et Cross-sectional Rural adolescents 1304 Beach Distance from the residence to Use of beach for PA Questionnairea OR (95%CI) of using the beach for PA in al. 2014, 12-15 years the closest beach using network those living > 800 m = 0.84 (0.51, 1.30) Australia analysis (≤ 800 m and > 800 m) in winter and 0.68 (0.50, 0.93) in summer Use of beach for PA (yes/no) Physical activity APARQ (MetMins/ OR of meeting PA guidelines if beach use week values obtained) reported = 1.45 (p-value=0.01) during winter and 3.6 (95%CI=1.9, 4.7) in summer Elliot et al. Cross-sectional Adults >16 years 71603 Seaside resort Questionnaire asking for visits Physical activity Questionnairea β (95%CI)=0.03 (0.02, 0.04) for seaside 2015, UK or other coast to seaside resorts or towns, or (duration of the visit visits and 0.02 (0.01, 0.04) for other coast visits other coast sites in the last week multiplied by METs visits using urban greenspace visits as associated to the reference. If only walkers included then activity performed numbers are: 0.07 (0.05, 0.08) and 0.10 during the visit) (0.07, 0.12), respectively Gilmer et al. Cross-sectional Youth 11-14 113 Coastal region Geographic regions based on Physical activity Youth Health Survey Mean (SD) of levels of PA between 2003, USA years (children of school location within the state regions: coastal 229 (3.68), piedmont 237 parents with of North Carolina (3.31), mountains 252 (2.99). Differences premature heart were significant (p<0.05) between coastal disease) and the other two regions Humpel et Cross-sectional Adults >40 years 399 Costal Based on postal code Neighbourhood Questionnaire Men: OR (95%CI)=1.69 (0.69, 4.18) al. 2004a, geographical walking Women: 3.32 (1.51, 7.29) Australia locations Exercise walking Questionnaire Men: 1.72 (0.70, 4.19) Women: 1.40 (0.62, 3.18) Pleasure walking Questionnaire Men: 1.59 (0.64, 3.95) Women: 1.65 (0.72, 3.82) Walking for Questionnaire Men: 0.94 (0.40, 2.19) commuting Women: 1.83 (0.87, 3.85)

Author Study design Age of the study N Blue space Tool to estimate blue space Outcome measured Tools to measure Main results (year, population exposure exposure outcome country) Humpel et Cross-sectional Adults 18-71 800 Costal Based on postal code Total walking IPAQ Men: OR (95%CI)=1.01 (0.64, 1.59) al. 2004b, years geographical Women: 1.00 (0.63, 1.58) Australia locations Total physical activity IPAQ Men: 1.12 (0.70, 1.76) Women: 0.99 (0.63, 1.57) Neighbourhood Questionnairea Men: 1.66 (1.04, 2.67) walking Women: 0.99 (0.62, 1.58) Karusisi et Cross-sectional Adults 30-79 7290b Water Geographic unit was used to Jogging behaviour in Questionnairea RR (95%CI) of jogging = 1.15 (1.03, al. 2012, years (fresh) calculate % of area covered with the previous week 1.36) France water in a buffer of 1km from (yes/no) the residence Perchoux et Cross-sectional Adults 30-79 4365 Lakes or Land-use maps to determine (Not reporting ) Questionnairea OR (95%CI) = 0.84 (0.71, 0.99) for al. 2015, years waterways presence of lakes/waterways for recreational walking residential BS, 0.83 (0.70, 0.98) for France different areas and buffers time within and residential + recreational BS, 0.96 (0.78, (residential – 1 km, work – 1 outside their 1.18) for all BS km, recreational – 500 m...) neighbourhood in the previous week White et al. Cross-sectional Adults 183755 Coast Land-cover map to determine Leisure and travel- Questionnairea OR (95%CI) for 5 days of 30 min of PA 2014, UK residential distance to the coast related physical or more/week=1.08 (1.03, 1.14) for (<1km, 1-5km, 5-20km, >20km) activity distance <1km vs >20km, when adjusted for coastal visits = 0.93 (0.89, 0.98) Coastal visits Questionnaire OR (95%CI)= 1.62 (1.55, 1.69) Wilson et al. Cross-sectional Adults 40-65 10286 River or coast MapInfo GIS (quintiles) to Time spent walking Questionnaire OR (95%CI) of walking ≥ 300 min = 2.06 2011, years define network distance from for recreation, exercise (1.41, 3.02) for those on Q1 (least Australia the residence to the river or the or to go to places in distance) vs those on Q5 (greatest coast the last week distance) Witten et al. Cross-sectional Adolescents and Between Beach GIS and land information New Sedentary behavior (< Questionnairea OR (95%CI) for those living further from 2008, New adults > 15 years 11233 and Zealand to obtain travel time to 30 min of PA in the the beach (> 31.8 min walking) = 1.04 Zealand 12425 access the nearest beach and past week) (0.81, 1.33) define access to (from the Meet the Questionnairea OR (95%CI) for those living further from neighbourhood) recommended level of the beach= 0.88 (0.76, 1.03) PA (at least 2.5 h on five or more days in the preceding week) Ying et al. Cross-sectional Adults 46-80 1100 River GIS to define 500m buffer PA Pedometer Coefficient (p-value)=-0.112 (0.231) 2015, China years distance to the river 904 APARQ: Adolescent Physical Activity Recall Questionnaire; BMI: body mass index; BS: blue space; IPAQ: International Physical Activity Questionnaire; PA: physical activity.

905 aNot validated questionnaire or not clarified if the questionnaire is validated. 906 bThe final number of subjects included in the analysis is not clear.

907 Table 4. Main characteristics and results of the studies on cardiovascular related outcomes and obesity. Author Study design Age of the study N Blue space Tool to estimate blue space Outcome measured Tools to measure Main results (year, population exposure exposure outcome country) Bergovec et Cross-sectional Adults 23-102 3054 Coastal Based on residence on a coastal Hypertensiona Physical OR (95%CI)=0.73 (0.59, 0.89) al. 2008, years (coronary hospitals (vs region (not explained how it is measurements Croatia heart disease continental) defined) Cholesterol and 0.90 (0.68, 1.20) for total cholesterol, patients) triglycerides 0.74 (0.60, 0.92) for decreased HDL, 1.01 (0.81, 0.943) for triglycerides Diabetes 1.86 (0.69, 1.06) Halonen et Longitudinal Adults (urban 25317 Lake, river, Land-cover map to determine BMI (normal weight, Self-reported height No associations and no patterns were al. 2014, (8 years of follow- areas) (15621 sea residential distance to each blue overweight and and weight observed between distance to blue spaces Finalnd up) nonmovers space (<250m to >750m) obesity) and BMI, neither among non-movers nor and 9696 among movers. The only statistically movers) significant association observed was for overweight vs normal weight in distance 500-750m vs <250m in non-movers [OR (95%CI)=1.24 (1.01, 1.52)] Kern et al. Cross-sectional Adults Not Coastal region Based on administrative Cardiovascular risk Not clearly reported Prevalence of high CVRB in the coastal 2009, reported (vs divisions of the regions of the burden (CVRB)b region=46.3% (95%CI 40.9, 51.8) for Croatia continental) country (Adriatic sea) men and 43.9% (41.4, 46.4) for women, and in the northern region=53.1 (46.7, 59.6) and 54.2 (49.5, 58.9; p<0.05), respectively Modesti et Cross-sectional Adults 15-69 10242 Coast Based on administrative Hypertension Physical OR (95%CI) = 1.53 (1.25, 1.84) al. 2013, years divisions of the regions of the measurements & self- compared to participants living in the Yemen country reported use of capital of the country antihypertensive drugs at time of interview Diabetes 0.88 (0.65, 1.20) Abdominal obesity 0.88 (0.72, 1.08) Qin et al. Cross-sectional Hypertensive 17656 Coastal county Based on administrative Overweight and Physical OR (95%CI)=1.10 (1.02, 1.18) for those 2013, China adults 45-75 (vs inland) divisions (two counties) obesity (BMI based) measurements living inland vs coastal years Abdominal obesity OR (95%)=1.53 (1.43, 1.64) for those (waist circumference living inland vs coastal based) Turek et al. Cross-sectional Adults 5840 Coast Based on residence in a coastal Hypertensiona Physical Difference with a p-value<0.001 (coastal 2001, region measurements Rijeka=20.6% and Split=29.2%

Author Study design Age of the study N Blue space Tool to estimate blue space Outcome measured Tools to measure Main results (year, population exposure exposure outcome country) Croatia prevalence vs non-coastal Osijek 23.8% and Zagreb 30.9% prevalence) Cholesterol, “No major differences between regions” triglycerides and lipoprotein BMI “No major differences between regions” Witten et al. Cross-sectional Adolescents and Between Beach GIS and land information New BMI Questionnairec β (95%CI) for those living further from 2008, New adults > 15 years 11233 and Zealand to obtain travel time to the beach= 0.13 (0.07, 0.18) Zealand 12425 access the nearest beach and define access to (from the neighbourhood) Wood et al. Ecological Children 10-11 1475617 Coast Euclidean distance between BMI > 95th percentile Physical β (95%CI) for those living nearer (< 1km) 2016, UK Cross-sectional years population-weighted centroid measurements vs further (> 20 km) = -0.684 (-1.141, - and the coast 0.227) Ying et al. Cross-sectional Adults 46-80 1100 River GIS to define 500m buffer BMI Physical Coefficient (p-value)=0.040 (0.16) 2015, China years distance to the river measurements Overweight/obesity Coefficient (p-value)=0.144 (0.022) Zhang et al. Cross-sectional Children 7-18 42296 Coastal Based on administrative Abdominal obesity Physical Prevalence in boys=coastal (17.71%) and 2014, China years districts (vs divisions (sixteen districts) (waist circumference measurements inland (14.97%), p<0.01 inland) based) Prevalence in girls= coastal (7.91%) and (6.72%), p<0.01 908 BMI: body mass index 909 aThey also measure plasma fibrinogen levels 910 bCVRB: variable created based on several cardiovascular related outcomes or determinants (cardiovascular related incidents, weight status, blood pressure, smoking status, physical activity 911 levels, alcohol consumption, nutrion). 912 cNot validated questionnaire or not clarified if the questionnaire is validated. 913

914 Figure 1. Selection procedure of the articles.

915

916 Figure 2. Summary (%) of the blue space type, environment of assessment and the method of evaluation used in the studies included by each 917 outcome assessed. 918

919 BS: blue space; CV: cardiovascular 920

Identification Medline Scopus Other sources 2438 2380 33

229 duplicates

Screening 4622 4531 excluded (title and abstract)

Eligibility 91 56 excluded (full text)

Finally included 35 Blue space Environment of Method of type assessment evaluation

General health (N=6)

Mental health & well-being (N=12)

Physical activity (N=13)

Obesity & CV related outcomes (N=10)

Non-inland Residential % of BS

Inland Use of BS Presence of BS in a buffer Both School Linear distance to a BS Network distance to a BS

Living in a coastal neighborhood Based postcode or living in a costal region Questionnaire asking about time spent in BS Others Supplemental Material

Blue spaces, human health and well-being: a systematic review

Mireia Gascon, Wilma Zijlema, Cristina Vert, Mathew P. White, Mark Nieuwenhuijsen

Table A. Additional characteristics of the studies included in the systematic review …………...2

Table B . Criteria for quality assessment of the studies …………...8

Table C . Specific scores for each item evaluated and the final quality scores and categories given to each study …………...9

Table A. Additional characteristics of the studies included in the systematic review. Author (year) Study design/ Statistical methods Co-variables of adjustment and interactions evaluated Other information population (N) Alcock et al. 2015 Longitudinal (18 Random effects V CAU level: income, employment and education deprivation and crime V It does not evaluate quality of BS years follow-up) linear regression V Individual level: age, education, marital status, living with children (<16y), V No mention of the minimal time of residence Adults (with within- and household income, work-limiting health, labour market status, residential V Green and other natural spaces evaluated as well (N=2020) between-individual accommodation, accommodation size, commuting time, data collection V Restricted to urban areas effects) wave V They take into account residential permanency Amoly et al. 2014 Cross-sectional Quasi-Poisson V CAU level: socioeconomic status V It does not evaluate quality of BS Children 7-10 years mixed effects V Individual level: gender, school level, ethnicity, preterm birth, V No mention of the minimal time of residence (N=2111) models breastfeeding, exposure to environmental tobacco smoke, maternal smoking V Green spaces evaluated as well during pregnancy, responding person, parental education, employment and V Restricted to urban areas marital status V Only 0.5% and 1.7% of the participants lived within 300 m and 500 m of the beach, respectively. Analyses finally not conducted and distance information not included to adjust for the association reported in Table 1. Ball et al. 2006 Cross-sectional Random effects V CAU level: aesthetics, safety, walking track length, street connectivity V It does not evaluate quality or use of BS Women 18-65 years logistic regression V Individual variables (statistically tested but not included): age, marital V No mention of the minimal time of residence (N=1282) status, presence of children in the home and pregnancy V PA information based on questionnaire V Individual variables included: education, family and friend’s social support, V Exposure information based on the neighbourhood club membership, dog owner, self-efficacy, enjoyment, barriers, intentions V Coarse assessment of coastal proximity and not defined based on actual residential distance Bauman et al. Cross-sectional Logistic regression V Individual level: age, sex, employment status, education level, country of V No mention of the minimal time of residence 1999 Adults birth V PA information based on questionnaire (N=16178) V Coarse assessment of coastal proximity and not defined based on actual residential distance Bergovec et al. Cross-sectional Logistic regression V Not clearly specified V Coarse assessment of coastal proximity and not defined based on 2008 Adults 23-102 years actual residential distance (N=3054) Brerenton et al. Cross-sectional Random effects V Area level: population density, congestion, commuting time, climatic and V Individual and GIS data linked at the electoral division level (not 2007 Adults >18 years linear regression environmental variables individual level) (N=1467) V Individual level: age, gender, employment status, educational attainment, V It does not evaluate quality or use of BS health, marital status, income, number of dependent children, household V No mention of the minimal time of residence tenure, Dublin

Author (year) Study design/ Statistical methods Co-variables of adjustment and interactions evaluated Other information population (N) De Vries et al. Cross-sectional Logistic regression V CAU level: urbanity V It does not evaluate quality or use of BS 2003 All ages (N=10197) V Individual level: age, gender, education, number of rooms, type of health V Participants have been living at least 1 year in the studied insurance, number of life-events residence V Interaction with education and urbanity degree V Exclusion of those with changes in urbanity in their neighbourhood V BS and health data collected at different moments with 9 years in between Edwards et al. Cross-sectional Logistic regression V Individual level: age, sex, SES, ethnicity V It does not evaluate quality BS 2014 Rural adolescents 12- V Presence of a major road (crossing busy road to go the beach) V No mention of the minimal time of residence 15 years V Stratification by season V PA information based on questionnaire (N=1304) V Information not based on exact home address but based on the nearest intersection Elliot et al. 2015 Cross-sectional Linear regression V Individual level: self-reported PA, age, gender, SES, work, marital status, V It does not evaluate quality BS Adults >16 years analysis ethnicity, disability V Comparison of energy expenditure between BS and GS (N=71603) V Visit characteristics: presence of other adults, children and dogs during the V Duration of the visit was asked visit, starting point of the visit, distance travelled to the destination, travel V Sensitivity analysis excluding dog-walking visits mode, ownership of a car, frequent green space users, seasonality and year V PA information based on questionnaire of the survey (2009-2014) V Stratification by urban/rural and by travel distance Gilmer et al. 2003 Cross-sectional ANCOVA V Individual level: age, gender, race, pubertal status, peer and family physical V This is a selected population of children with at least one parent Youth 11-14 years activity, SES and education of the parents with a history of premature coronary heart disease (N=113) V Community characteristics: SES and geographic area V PA information based on questionnaire V Stratification by peer physical activity V No mention of the minimal time of residence V Coarse assessment of coastal proximity and not defined based on actual residential distance Halonen et al. Longitudinal Logistic regression V Individual level: age (at baseline), sex, education (as a proxy for SES), V The cohort is based on workers of the public administration 2014 (8 years of follow- chronic somatic illnesses, smoking, alcohol use and physical activity V Height and weight self-reported up) V Neighbourhood level: socioeconomic disadvantage V Does not include green spaces in the analysis Adults V Separate analysis for movers and non-movers (N=25317) V It does not evaluate use or quality BS

Author (year) Study design/ Statistical methods Co-variables of adjustment and interactions evaluated Other information population (N) Huynh et al. 2013 Cross-sectional Logistic regression V Area level: neighbourhood aesthetics and SES V It does not evaluate quality or use of BS Children 11-16 years V Individual level: SES, perceived neighbourhood safety V No mention of the minimal time of residence (N=17249) V Stratification by rural, small city, metropolitan area V Students living > 1h travel distance from school excluded V Effect modifiers: age, gender, ethnicity, urban/rural V 5 km buffer as a proxy for the neighbourhood environment (all children from same school had the same exposure estimates) Humpel et al. Cross-sectional Logistic regression V Individual level: age, education, physical environment attributes (perceived V Sample of clients of a health insurance organization 2004a Adults >40 years environment) V PA information based on questionnaire (N=399) V Stratified by gender V No mention of the minimal time of residence V Low response rate study V It does not evaluate use of BS V Coarse assessment of coastal proximity and not defined based on actual residential distance Humpel et al. Cross-sectional Logistic regression V Individual level: age, education, physical environment attributes (perceived V Sample of university staff members 2004b Adults 18-71 years environment) V PA information based on questionnaire (N=800) V Stratified by gender V No mention of the minimal time of residence V It does not evaluate use of BS V Coarse assessment of coastal proximity and not defined based on actual residential distance Karusisi et al. Cross-sectional Log binomial V Neighbourhood level (1km buffer): SES, physical, service related, socio- V It does not evaluate quality or use of BS 2012 Adults 30-79 years regression interactional and symbolic environment, % GS and parks V No mention of the minimal time of residence (N=7290) V Individual level: age, sex, education, marital status, occupation, household V Daily meteorological data collected for the day before of the income, homeownership, perceived financial strain, human development recruitment and 7 days later index of the country of birth, energy expenditure at work over the previous V PA (jogging) information based on questionnaire week, proximity of friends and family in the neighbourhood Kern et al. 2009 Cross-sectional Prevalence V No variables of adjustment V It does not evaluate quality or use of BS Adults (descriptive V No mention of the minimal time of residence (N=3479588) statistics) V Coarse assessment of coastal proximity and not defined based on distance MacKerron et al. Longitudinal Regression analyses V Where the participants is (indoors, in a vehicle, outdoors), current activity V Sample limited to sample Iphone users 2013 Adults with cluster-robust (at home, at work, elsewhere), weather conditions when outdoors, selected V Self-selection of the participants (or moment of participation) (N=21947) sandwich estimator activities, companionship (children, spouse, etc) V It does not evaluate the quality of BS for correlated V Objective measurement of use of BS standard errors

Author (year) Study design/ Statistical methods Co-variables of adjustment and interactions evaluated Other information population (N) Modesti et al. Cross-sectional Logistic regression V Age, gender, urban or rural location, level of education, average air V It does not evaluate quality or use of BS 2013 Adults 15-69 years temperature at the time of surveys, self-reported sedentary lifestyle, V No mention of the minimal time of residence (N=10242) smoking habits, vegetable and fruit consumption, blood glucose, high V Coarse assessment of coastal proximity and not defined based on cholesterol and high triglycerides actual residential distance Nutsford et al. Cross-sectional Linear regression V CAU level: SES, population density, crime rates, housing quality index V It does not evaluate quality or use of BS 2016 Adolescents and (HQI) V No mention of the minimal time of residence adults > 15 years V Individual level: sex, age, income V VVI is an index that takes into account slope, aspect, distance and (N= 442) V Others: missing teeth (SES proxy) as an outcome (test) elevation of visible areas relative to the observer location V Exposure information not based on exact home address but the CAU Perchoux et al. Cross-sectional Zero-inflated V Individual level: age, sex, education, employment status, household income, V It does not evaluate quality or use of BS 2015 Adults 30-79 years negative binomial living with at least 1 child <14 years, location in the regions, citizenship V Did not take into account length of residence (N=4365) models V Different environments of activity evaluated Qin et al. 2013 Hypertensive adults Logistic regression V Sex, smoking, alcohol drinking, antihypertensive treatment status, V It does not evaluate quality or use of BS 45-75 years hypertension grade, standard of living, meat, fruit and vegetable V No mention of the minimal time of residence (N=17656) consumption, education level, physical activity, family history of V Coarse assessment of coastal proximity and not defined based on hypertension, diabetis, coronary heart disease or stroke distance Rogerson et al. Pre/post ANOVA V No adjustment for any co-variable V Non-nature-based place for the comparison 2016 observational study V Data collected on four different days between September and Adult runners November Triguero-Mas et Cross-sectional Logistic regression V CAU level: indicators neighbourhood socioeconomic status V It does not evaluate quality or use of BS al. 2015 34-64 years V Individual level: gender, age, education level, birth place, type of health V No mention of the minimal time of residence (N=8793) insurance, marital status and indicators of household status V Stratified: gender, SES, degree of urbanization Turek et al. 2001 Cross-sectional X2 test, ANOVA V No adjustment for any co-variable V It does not evaluate quality or use of BS Adults V No mention of the minimal time of residence (N=5840) V Coarse assessment of coastal proximity and not defined based on actual residential distance Wheeler et al. Ecological cross- Linear regression V Outcome sex and age standardized (% at CAU) V It does not evaluate quality or use of BS 2012 sectional V CAU level: deprivation indices (income, employment, education, crime, V No mention of the minimal time of residence Not specified environment) and % of GS N=32482 (CAUs) V Stratified by income deprivation, urban/rural

Author (year) Study design/ Statistical methods Co-variables of adjustment and interactions evaluated Other information population (N) Wheeler et al. Ecological cross- Linear regression V Outcome sex and age standardized (% at CAU) V It does not evaluate quality or use of BS 2015 sectional V CAU level: income, education and employment status and urban/rural V No mention of the minimal time of residence Not specified V Mutual adjustment for all other natural land cover types (excluding N=31672 (CAUs) urban) White et al. 2013a Longitudinal Linear regression V Individual level: age, education, marital status, income, living with V It does not evaluate quality or use of BS Adults > 25 years children, work-limiting health status, labour market status, residence type, V No mention of the minimal time of residence (N=15361) household space, commute length, green space White et al. 2013b Longitudinal Linear regression V CAU level: employment, education, income and crime, % of GS V It does not evaluate quality or use of BS Adults V Individual level: age, education, marital status, living with children, work- V Does not separate between movers and non-movers along time (N=10168 to 12818) limiting health status, labour status, commuting, household residence, income and space White et al. 2013c Cross-sectional Linear regression V CAU level: employment, education, income and crime, % of GS and % of V It does not evaluate quality or use of BS > 16 years freshwater V No mention of the minimal time of residence (N=4255) V Individual level: gender, age, SES, main activity engaged, visit duration, visit companions, starting location, distance travelled, mode of travel

White et al. 2014 Cross-sectional Logistic regression V Individual level: gender, age, occupational social grade, employment status, V PA information based on questionnaire Adults marital status, number of children in the household, ethnicity, work limiting V It does not evaluate quality of BS (N=183755) health status, car access, dog ownership. Also included season and year of data collection V CAU level: amount of greenness, local area deprivation (unemployment and crime), % of GS V Stratified by region and mediation analysis including coast visits Wilson et al. 2011 Cross-sectional Multinomial logit V Neighbourhood level: socioeconomic disadvantage V It does not evaluate quality or use of BS Adults 40-65 years models V Individual level: age, sex, education, occupation, living arrangement, V No mention of the minimal time of residence (N=10286) household income V Multiple testing Witten et al. 2008 Cross-sectional Linear and logistic V CAU levels: stratum (ethnic composition of the meshblock), deciles of V It does not evaluate quality or use of BS Adults >15 years regressions number of respondents in the meshblock, area deprivation, urbanity V No mention of the minimal time of residence (N=12529) V Individual-level: age, sex, education, social class, receipt of benefits, working/not working and household income, number of adults in the household and ethnicity Wood et al. 2016 Ecological Linear regression V Area level: % of greenspace, index of multiple deprivation, % of population V It does not evaluate quality or use of BS Cross-sectional 10-14 years, % of population of non-white ethnicity, rural/urban V No mention of the minimal time of residence Children 10-11 years V Stratified by rural/urban V Coarse assessment of coastal proximity and not defined based on

Author (year) Study design/ Statistical methods Co-variables of adjustment and interactions evaluated Other information population (N) (N=1475617) actual residential distance Ying et al. 2015 Cross-sectional Hierarchical V Individual-level: age, gender, education, employment. Land-use mix, net V It does not evaluate quality or use of BS Adults 46-80 years multilevel models residential density, street connectivity, proximity to parkland and square, V No mention of the minimal time of residence (N=1100) road areas Zhang et al. 2014 Cross-sectional Prevalence V No variables of adjustment V It does not evaluate quality or use of BS Children 7-18 years (descriptive V No mention of the minimal time of residence (N=42296) statistics) V Coarse assessment of coastal proximity and not defined based on distance ANOVA: analysis of variance; ANCOVA: analysis of covariance; BS: blue spaces; CAU: Census at Unit level; GS: green spaces; PA: physical activity; SES: socioeconomic status

Table B. Criteria for quality assessment of the studies

Study design 0 = ecological, 1 = cross-sectional or case-control study, 2 = longitudinal or intervention study

Confounding factors 0 = no confounding factors considered, 1 = confounding factors considered but some key confounders omitted, 2 = careful consideration of confounders

Statistics 0 = flaws in or inappropriate statistical testing or interpretation of statistical tests that may have affected results, 1 = appropriate statistical testing and interpretation of tests

Potential bias 0 = other study design or conduct issues that may have led to bias, 1 = no other serious study flaws

Multiplicity 0 = exposure of interest one of the many variables being tested, 1 = exposure of interest the main variable tested

Outcome assessment 0 = self-reported non-validated questionnaires, 1 = self-reported validated questionnaires, 2 = interviews conducted by experts or clinical records, or physical measurements

Blue exposure assessment 0 = self-reported based or based on regional residence, 1 = satellite system or land-cover map based

Use of blue space 0 = not measured and/or not included in the analysis, 1=measured and included in the analysis

Effect size 0 = incomplete information, 1 = complete information (estimate and standard error or confidence interval)

Participants have been living at 0 = no or not clearly specified, 1= yes least 1 year in the studied area

Table C. Specific scores for each item evaluated and the final quality scores and categories given to each study

Time of Study Confounding Potential Outcome BS Use Effect Statistics Multiplicity residence Score Score Quality design factors bias assessment assessment of BS size (0-1) (0-1) considered (absolute) (%) a category (0-2) (0-2) (0-1) (0-2) (0-1) (0-1) (0-1) (0-1) Alcock et al. 2015 2 2 1 1 1 1 1 0 1 1 11 85 Excellent Amoly et al. 2014 1 2 1 1 1 1 0 1 1 0 9 69 Good Ball et al. 2007 1 2 1 1 1 1 1 0 1 0 9 69 Good Bauman et al. 1999 1 1 1 1 1 1 1 0 1 0 8 62 Good Bergovec et al. 2008 1 0 1 1 0 2 0 0 1 0 6 46 Fair Brereton et al. 2008 1 2 1 1 1 0 1 0 1 0 8 62 Good De Vries et al. 2003 1 1 1 0 0 1 1 0 1 1 7 54 Fair Edwards et al. 2014 1 1 1 1 1 1 1 1 0 0 8 62 Good Elliott et al. 2015 1 2 1 1 1 0 0 1 1 NA 8 67 Good Gilmer et al. 2003 1 2 0 0 1 1 0 0 0 0 5 38 Poor Halonen et al. 2014 2 2 1 1 1 0 1 0 1 1 10 77 Good Humpel et al. 2004a 1 1 1 1 1 0 1 0 1 0 7 54 Fair Humpel et al. 2004b 1 1 1 1 1 1 1 0 1 0 8 62 Good Huynh et al. 2013 1 2 1 0 1 1 1 0 1 0 8 62 Good Karusisi et al. 2012 1 2 1 1 1 0 1 0 1 0 8 62 Good Kern et al. 2009 1 0 0 1 1 0b 0 0 1 0 4 31 Poor MacKerron et al. 2013 2 1 1 0 1 0 1 1 1 NA 8 67 Good Modesti et al. 2013 1 2 1 1 0 2 0 0 1 0 8 62 Good Nutsford et al. 2016 1 2 1 0 1 1 1 0 1 0 8 62 Good Perchoux et al. 2015 1 1 1 1 0 0 1 0 1 0 6 46 Fair Qin et al. 2013 1 2 1 1 0 2 0 0 1 0 8 62 Good Rogerson et al. 2016 0c 0 0 0 1 1 NA 0 1 NA 3 27 Poor Triguero-Mas et al. 2015 1 2 1 1 0 1 1 0 1 0 8 62 Good

Time of Study Confounding Potential Outcome BS Use Effect Statistics Multiplicity residence Score Score Quality design factors bias assessment assessment of BS size (0-1) (0-1) considered (absolute) (%) a category (0-2) (0-2) (0-1) (0-2) (0-1) (0-1) (0-1) (0-1) Turek et al. 2000 1 0 0 1 0 2 0 0 0 0 4 31 Poor Wheeler et al. 2012 0 1 1 0 1 0 1 0 1 0 5 38 Poor Wheeler et al. 2015 0 1 1 0 0 0 1 0 1 0 4 31 Poor White et al. 2013a 2 2 1 1 1 1 1 0 1 0 10 77 Good White et al. 2013b 2 2 1 1 1 1 1 0 1 0 10 77 Good White et al. 2013c 1 2 1 1 1 0 0 1 1 0 8 62 Good White et al. 2014 1 2 1 1 1 0 1 1 1 0 9 69 Good Wilson et al. 2011 1 2 1 1 0 1 1 0 1 0 8 62 Good Witten et al. 2008 1 2 1 1 1 0 1 0 1 0 8 62 Good Wood et al. 2016 0 1 1 0 1 2 1 0 1 0 7 54 Fair Ying et al. 2015 1 2 1 1 1 2 1 0 1 0 10 77 Good Zhang et al. 2014 1 0 0 1 1 2 0 0 1 0 6 46 Fair NA=not applicable aFor each study the total score was calculated by adding the scores of the 10 dimensions defined and expressing them as a percentage of the maximum score (for some studies some of the items did not apply to be evaluated (NA) so the maximum score could vary among studies). Based on the score in %, the following quality scores were obtained: Excellent quality (score ≥81%), good quality (between 61 and 80%), fair quality (between 41 and 60%), poor quality (between 21 and 40%) and very poor quality (≤20%) bKern et al. gets a “0” for outcome assessment because the methods for outcome measurements are not clearly reported. cRogerson et al. gets a “0” for study design because it is a pre-/post-natural intervention study, without controls and several flaws in the design.